%7i dZddlmZmZddlmZmZmZmZmZm Z m Z m Z ddl m Z Gdde ZGdde ZGd d e ZGd d e ZGd de ZGdde Zy)zr Hyper Forecast Learning System - Database Models Models for logging predictions and tracking calibration history )datetimedate)ColumnIntegerStringDateTimeDateNumericTextJSON)Basec`eZdZdZdZeedZeede jZ eeddZ ee dd d Zee dd d Zee d d d ZeediZee dd d Zeed d ZeeddZeed ZdZy) ForecastLoga] Logs every prediction made by the Hyper Forecast engine. This is the CRITICAL table for the learning system. Flow: 1. Engine makes prediction -> saves entry with valor_previsto 2. Daily job runs at 03:00 -> fills valor_real from actual orders 3. Weekly job analyzes -> uses erro_percentual to calibrate multipliers forecast_logsT primary_keyFnullabledefaultrindex r2 heuristic_v1rNc pd|jd|jd|jd|jd S)Nz )id hora_alvovalor_previsto valor_realselfs J/var/www/hypershopcomercio.com.br/hyper-ai/app/models/forecast_learning.py__repr__zForecastLog.__repr__4s=twwir$..)9@S@S?TTZ[_[j[jZkklmmN)__name__ __module__ __qualname____doc__ __tablename__rrr$rrutcnowtimestamp_previsaor%r r&r'erro_percentualr fatores_usadosbaseline_usador modelo_versao calibratedcalibration_impactr+r,r*rr s$M T *B 5(//Rx%tMq 3/Jt4nr,rcJeZdZdZdZeedZeede jZ ee dddZ ee d d Zeed d d Zeed d d Zeed d d Zeed Zeed d d Zeed ZdZy)CalibrationHistoryz Tracks all automatic calibration adjustments made to multipliers. Provides audit trail and allows reverting if needed. calibration_historyTrFrrrdrrrc pd|jd|jd|jd|jd S)Nz r#) tipo_fator fator_chavevalor_anterior valor_novor(s r*r+zCalibrationHistory.__repr__^s@%doo%6a8H8H7IDL_L_K``deietetduuvwwr,N)r-r.r/r0r1rrr$rrr2data_calibracaorrBrCr rDrE erro_medioamostrasajuste_percentualr notasr+r:r,r*r<r<8s*M T *BXxOOr U$?Ju5KGAqME:N1 6J1 6Jg.Hwq!}t< 4$ 'Exr,r<c0eZdZdZdZeedZeedddZ eedd Z ee d d dd Z eedZ eeej ej ZeeddZeeddZddifZdZy)MultiplierConfigz Stores current multiplier values that can be auto-updated. This replaces hardcoded constants with database-driven values. multiplier_configTrrFrr>rrr??rrronupdaterrr schemaNcVd|jd|jd|jdS)NzEw+I8X__xWMvbz95IF1Is +F 4N Kr,rLceZdZdZdZeedZeedddZ eedZ ee d d d Z ee d d d Z ee d d d Zee d d dZee d d dZeeiZeedZeegZeedd Zeedd Zeeej4ZdZy)LearningSnapshotz Daily snapshot of learning metrics for historical analysis. Created at 23:55 each day to capture the day's performance. learning_snapshotsTrF)runiquerrrrrr r>c<d|jd|jdS)Nz)dataacuraciar(s r*r+zLearningSnapshot.__repr__s#DII;fT]]O2FFr,N)r-r.r/r0r1rrr$r rdtotal_previsoesr rGerro_absoluto_mediorereceita_prevista_totalreceita_real_totalr fatores_performanceajustes_realizadosdetalhes_ajustesr melhor_fator pior_fatorrrr2 created_atr+r:r,r*r^r^s)M T *B $t4 @DWa0O1 5J A>gamd3H$GBNA>A:!r2 3dB/&+5Ls d3J(//:JGr,r^cveZdZdZdZeedZeeddZ ee dZ ee ddZ ee dZeed d dd Zeed dZeeddZeeddZeeej,Zeeej,ej,ZdZy) ForecastEventu{ Special events that affect sales predictions. Examples: Black Friday, Natal, Dia das Mães, promoções, etc. forecast_eventsTrr>FrrrrNrrmanualrrr YrOc pd|jd|jd|jd|jd S)Nz)nome data_iniciodata_fim multiplicadorr(s r*r+zForecastEvent.__repr__s<  2d.>.>-?s4==/QSTXTfTfSggjkkr,N)r-r.r/r0r1rrr$rrwr descricaor rxryr rzrU recorrenteativorrr2ro updated_atr+r:r,r*rqrqs&M T *B &+ .Dtd+IT:KdU+H71a=5#FM &*h /Dq 3/J 6!9c *E(//:J(//HOOTJlr,rqceZdZdZdZeedZeedddZ eeddZ eed dZ eed d Z ee ej Zd Zy) AllowedFactorzu Whitelist for allowed factor keys to prevent garbage data. E.g. factor_type='momentum', factor_key='up' allowed_factorsTrrFrrrrurc<d|jd|jdS)Nzrst$SSS +n$+n\'x'xT$Kt$KN%Gt%GP"lD"lJGDGr,